Skip to contentSkip to conversation

Add framework for passing secrets to QEMU

PATCH Superseded Daniel P. Berrange 2 participants 13 messages Nov 24, 2015 Crypto API

v2 of 4 Superseded Latest is v4 →

lore
Share & tools
lore
#0 lore raw

v2 cover Add framework for passing secrets to QEMU

This small series contains just the first two patches for adding the secrets object to QEMU previously shown here:

https://lists.gnu.org/archive/html/qemu-devel/2015-11/msg04748.html

The QCryptoSecret object provides a QOM object that allows passing secrets (passwords, encryption keys, etc) to QEMU in a secure manner, via an external file, or on the CLI or monitor with encryption. It also uses this to allow use of encrypted x509 keys with the TLS handling code.

Changed in v2:

- Fix version tag in QAPI schema to be 2.6 (Eric)
- Changed "echo -n" to "printf" (Eric)
- Misc typo fixes (Eric)
- Added a genmeric qbase64_decode() wrapper around g_base64_decode() that does error checking (Markus)
- Convert callers of g_base64_decode() to qbase64_decode() to get error checking (Markus)

Daniel P. Berrange (5):
util: add base64 decoding function
qemu-char: convert to use error checked base64 decode
qga: convert to use error checked base64 decode
crypto: add QCryptoSecret object class for password/key handling
crypto: add support for loading encrypted x509 keys

 crypto/Makefile.objs          |   1 +
 crypto/secret.c               | 540 ++++++++++++++++++++++++++++++++++++++++++
 crypto/tlscredsx509.c         |  47 ++++
 include/crypto/secret.h       | 148 ++++++++++++
 include/crypto/tlscredsx509.h |   1 +
 include/qemu/base64.h         |  56 +++++
 qapi-schema.json              |   2 -
 qapi/crypto.json              |  14 ++
 qemu-char.c                   |   8 +-
 qemu-options.hx               |  85 ++++++-
 qga/commands-posix.c          |  11 +-
 qga/commands-win32.c          |  11 +-
 qga/commands.c                |  13 +-
 qmp-commands.hx               |   2 -
 tests/.gitignore              |   2 +
 tests/Makefile                |   5 +
 tests/test-base64.c           |  97 ++++++++
 tests/test-crypto-secret.c    | 446 ++++++++++++++++++++++++++++++++++
 util/Makefile.objs            |   1 +
 util/base64.c                 |  60 +++++
 20 files changed, 1539 insertions(+), 11 deletions(-)
 create mode 100644 crypto/secret.c
 create mode 100644 include/crypto/secret.h
 create mode 100644 include/qemu/base64.h
 create mode 100644 tests/test-base64.c
 create mode 100644 tests/test-crypto-secret.c
 create mode 100644 util/base64.c
Signature · 1 line
2.5.0
#1 lore raw

PATCH v2 1/5 util: add base64 decoding function

The standard glib provided g_base64_decode doesn't provide any kind of sensible error checking on its input. Add a QEMU custom wrapper qbase64_decode which can be used with untrustworthy input that can contain invalid base64 characters, embedded NUL characters, or not be NUL terminated at all.

---
 include/qemu/base64.h | 56 +++++++++++++++++++++++++++++
 tests/.gitignore      |  1 +
 tests/Makefile        |  3 ++
 tests/test-base64.c   | 97 +++++++++++++++++++++++++++++++++++++++++++++++++++
 util/Makefile.objs    |  1 +
 util/base64.c         | 60 +++++++++++++++++++++++++++++++
 6 files changed, 218 insertions(+)
 create mode 100644 include/qemu/base64.h
 create mode 100644 tests/test-base64.c
 create mode 100644 util/base64.c
include/qemu/base64.h new file +56 -0
new file mode 100644
index 0000000..34f44d9
@@ -0,0 +1,56 @@
+/*
+ * QEMU base64 helpers
+ *
+ * Copyright (c) 2015 Red Hat, Inc.
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This library is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, see <http://www.gnu.org/licenses/>.
+ *
+ */
+
+#ifndef QEMU_BASE64_H__
+#define QEMU_BASE64_H__
+
+#include "qemu-common.h"
+
+
+/**
+ * qbase64_decode:
+ * @input: the (possibly) base64 encoded text
+ * @in_len: length of @input or -1 if NUL terminated
+ * @out_len: filled with length of decoded data
+ * @errpr: pointer to uninitialized error object
+ *
+ * Attempt to decode the (possibly) base64 encoded
+ * text provided in @input. If the @input text may
+ * contain embedded NUL characters, or may not be
+ * NUL terminated, then @in_len must be set to the
+ * known size of the @input buffer.
+ *
+ * Note that embedded NULs, or lack of a NUL terminator
+ * are considered invalid base64 data and errors
+ * will be reported to this effect.
+ *
+ * If decoding is successful, the decoded data will
+ * be returned and @out_len set to indicate the
+ * number of bytes in the decoded data.
+ *
+ * Returns: the decoded data or NULL
+ */
+uint8_t *qbase64_decode(const char *input,
+                        size_t in_len,
+                        size_t *out_len,
+                        Error **errp);
+
+
+#endif /* QEMU_BUFFER_H__ */
tests/.gitignore +1 -0
index 1e55722..00a7fed 100644
@@ -8,6 +8,7 @@ check-qom-interface
 check-qom-proplist
 rcutorture
 test-aio
+test-base64
 test-bitops
 test-blockjob-txn
 test-coroutine
tests/Makefile +3 -0
index b937984..0e1289f 100644
@@ -84,6 +84,7 @@ check-unit-$(CONFIG_GNUTLS) += tests/test-crypto-tlscredsx509$(EXESUF)
 check-unit-$(CONFIG_GNUTLS) += tests/test-crypto-tlssession$(EXESUF)
 check-unit-$(CONFIG_LINUX) += tests/test-qga$(EXESUF)
 check-unit-y += tests/test-timed-average$(EXESUF)
+check-unit-y += tests/test-base64$(EXESUF)
 
 check-block-$(CONFIG_POSIX) += tests/qemu-iotests-quick.sh
 
@@ -417,6 +418,8 @@ tests/test-vmstate$(EXESUF): tests/test-vmstate.o \
 tests/test-timed-average$(EXESUF): tests/test-timed-average.o qemu-timer.o \
 	libqemuutil.a stubs/clock-warp.o stubs/cpu-get-icount.o \
 	stubs/notify-event.o stubs/replay.o
+tests/test-base64$(EXESUF): tests/test-base64.o \
+	libqemuutil.a libqemustub.a
 
 tests/test-qapi-types.c tests/test-qapi-types.h :\
 $(SRC_PATH)/tests/qapi-schema/qapi-schema-test.json $(SRC_PATH)/scripts/qapi-types.py $(qapi-py)
tests/test-base64.c new file +97 -0

97 lines not on this page · raw message

util/Makefile.objs +1 -0
index 89dd80e..8620a80 100644
@@ -30,3 +30,4 @@ util-obj-y += qemu-coroutine-sleep.o
 util-obj-y += coroutine-$(CONFIG_COROUTINE_BACKEND).o
 util-obj-y += buffer.o
 util-obj-y += timed-average.o
+util-obj-y += base64.o
util/base64.c new file +60 -0

60 lines not on this page · raw message

Signature · 1 line
2.5.0
#2 lore raw

Re: [Qemu-devel] [PATCH v2 1/5] util: add base64 decoding function

Quoted 7 lines · Daniel P. Berrange

The standard glib provided g_base64_decode doesn't provide any kind of sensible error checking on its input. Add a QEMU custom wrapper qbase64_decode which can be used with untrustworthy input that can contain invalid base64 characters, embedded NUL characters, or not be NUL terminated at all.

Quoted 8 lines

---

+/**
+ * qbase64_decode:
+ * @input: the (possibly) base64 encoded text
+ * @in_len: length of @input or -1 if NUL terminated
+ * @out_len: filled with length of decoded data
+ * @errpr: pointer to uninitialized error object

s/errpr/errp/

include/qemu/base64.h line 54

Quoted 7 lines

+ * Returns: the decoded data or NULL
+ */

include/qemu/base64.h new file +5 -0
+uint8_t *qbase64_decode(const char *input,
+                        size_t in_len,
+                        size_t *out_len,
+                        Error **errp);
+

Is char* any easier to work with than uint8_t* as the return type? I'm fine with either, and actually like that uint8_t doesn't cause unintentional sign-extension on 8-bit input, but just want to make sure we aren't forcing the majority of our callers to cast back to a more convenient type.

Quoted 7 lines
tests/test-base64.c new file +7 -0
+static void test_base64_good(void)
+{
+    const char *input = "QmVjYXVzZSB3ZSBmb2N1c2VkIG9uIHRoZS"
+        "BzbmFrZSwgd2UgbWlzc2VkIHRoZSBzY29ycGlvbi4=";
+    const char *expect = "Because we focused on the snake, "
+        "we missed the scorpion.";
+

tests/test-base64.c line 42

Quoted 10 lines
tests/test-base64.c new file +10 -0
+    size_t len;
+    uint8_t *actual = qbase64_decode(input,
+                                     -1,
+                                     &len,
+                                     &error_abort);
+
+    g_assert(actual != NULL);
+    g_assert_cmpint(len, ==, strlen(expect));
+    g_assert_cmpstr((char *)actual, ==, expect);
+    g_free(actual);

For example, this demonstrates a caller having to cast. But it doesn't show whether more callers want char* or uint8_t*.

Quoted 7 lines
tests/test-base64.c new file +7 -0
+
+static void test_base64_embedded_nul(void)
+{
+    const char input[] = "There's no such\0thing as a free lunch.";
+
+    test_base64_bad(input, G_N_ELEMENTS(input) - 1);
+}

tests/test-base64.c line 78

Quoted 10 lines
tests/test-base64.c new file +10 -0
+
+
+static void test_base64_not_nul_terminated(void)
+{
+    char input[] = "There's no such\0thing as a free lunch.";
+    input[G_N_ELEMENTS(input) - 1] = '!';
+
+    test_base64_bad(input, G_N_ELEMENTS(input) - 1);
+}
+

Did you mean to have an embedded NUL in the second test, or should you change that \0 to space?

util/base64.c line 26

Quoted 6 lines

+++ b/util/base64.c

util/base64.c new file +4 -0
+#include "qemu/base64.h"
+
+static const char *base64_valid_chars =
+    "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/=";

Do we want to allow newlines (perhaps by adding a bool parameter)? After all, although newline is not valid in base64, it is the one character that GNU coreutils special-cases (produce on output every --wrap columns, ignore on input without needing --ignore-garbage) to make base64 blocks easier to read by breaking into lines rather than one long string - and which may be relevant if someone is pasting output from base64(1) into QMP.

Signature · 2 lines
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org
#3 lore raw

Re: [Qemu-devel] [PATCH v2 1/5] util: add base64 decoding function

Quoted 19 lines · Eric Blake
Quoted 7 lines · Daniel P. Berrange

The standard glib provided g_base64_decode doesn't provide any kind of sensible error checking on its input. Add a QEMU custom wrapper qbase64_decode which can be used with untrustworthy input that can contain invalid base64 characters, embedded NUL characters, or not be NUL terminated at all.

Quoted 8 lines

---

+/**
+ * qbase64_decode:
+ * @input: the (possibly) base64 encoded text
+ * @in_len: length of @input or -1 if NUL terminated
+ * @out_len: filled with length of decoded data
+ * @errpr: pointer to uninitialized error object

s/errpr/errp/

Quoted 13 lines

include/qemu/base64.h line 54

Quoted 7 lines

+ * Returns: the decoded data or NULL
+ */

include/qemu/base64.h new file +5 -0
+uint8_t *qbase64_decode(const char *input,
+                        size_t in_len,
+                        size_t *out_len,
+                        Error **errp);
+

Is char* any easier to work with than uint8_t* as the return type? I'm fine with either, and actually like that uint8_t doesn't cause unintentional sign-extension on 8-bit input, but just want to make sure we aren't forcing the majority of our callers to cast back to a more convenient type.

I think using 'char *' encourages dangerous behaviour as you should not assume that the decoded data is a NUL terminated string without embedded NULs. It is arbirary binary data, which should be sanitized before using as a regular C 'char *' string. So uint8 forces the callers to thing about this. Fortunately all callers do currently handle this correctly, since g_base64_decode also returns a uint8_t equivalent type.

Quoted 12 lines
Quoted 12 lines
tests/test-base64.c new file +12 -0
+
+static void test_base64_embedded_nul(void)
+{
+    const char input[] = "There's no such\0thing as a free lunch.";
+
+    test_base64_bad(input, G_N_ELEMENTS(input) - 1);
+}
+
+
+static void test_base64_not_nul_terminated(void)
+{
+    char input[] = "There's no such\0thing as a free lunch.";
Quoted 8 lines

tests/test-base64.c line 78

Quoted 5 lines
tests/test-base64.c new file +5 -0
+    input[G_N_ELEMENTS(input) - 1] = '!';
+
+    test_base64_bad(input, G_N_ELEMENTS(input) - 1);
+}
+

Did you mean to have an embedded NUL in the second test, or should you change that \0 to space?

Sigh yes, there shouldn't be the embedded NUL in the second test.

Quoted 5 lines

util/base64.c line 26

Quoted 4 lines
util/base64.c new file +4 -0
+#include "qemu/base64.h"
+
+static const char *base64_valid_chars =
+    "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/=";
Quoted 7 lines

Do we want to allow newlines (perhaps by adding a bool parameter)? After all, although newline is not valid in base64, it is the one character that GNU coreutils special-cases (produce on output every --wrap columns, ignore on input without needing --ignore-garbage) to make base64 blocks easier to read by breaking into lines rather than one long string - and which may be relevant if someone is pasting output from base64(1) into QMP.

I'll test that glib correctly decodes base64 with newlines. Assuming it does then we should definitely allow it and extend the unit test to cover it too.

Regards,
Daniel

Signature · 4 lines
|: http://berrange.com      -o-    http://www.flickr.com/photos/dberrange/ :|
|: http://libvirt.org              -o-             http://virt-manager.org :|
|: http://autobuild.org       -o-         http://search.cpan.org/~danberr/ :|
|: http://entangle-photo.org       -o-       http://live.gnome.org/gtk-vnc :|
#4 lore raw

PATCH v2 2/5 qemu-char: convert to use error checked base64 decode

In reply to Daniel P. Berrange

Switch from using g_base64_decode over to qbase64_decode in order to get error checking of the base64 input data.

---
 qapi-schema.json | 2 --
 qemu-char.c      | 8 +++++++-
 qmp-commands.hx  | 2 --
 3 files changed, 7 insertions(+), 5 deletions(-)
qapi-schema.json +0 -2
index 8b1a423..7d92c04 100644
@@ -295,8 +295,6 @@
 # @format: #optional data encoding (default 'utf8').
 #          - base64: data must be base64 encoded text.  Its binary
 #            decoding gets written.
-#            Bug: invalid base64 is currently not rejected.
-#            Whitespace *is* invalid.
 #          - utf8: data's UTF-8 encoding is written
 #          - data itself is always Unicode regardless of format, like
 #            any other string.
qemu-char.c +7 -1
index 5448b0f..ac01d73 100644
@@ -32,6 +32,7 @@
 #include "qapi/qmp-input-visitor.h"
 #include "qapi/qmp-output-visitor.h"
 #include "qapi-visit.h"
+#include "qemu/base64.h"
 
 #include <unistd.h>
 #include <fcntl.h>
@@ -3259,7 +3260,12 @@ void qmp_ringbuf_write(const char *device, const char *data,
     }
 
     if (has_format && (format == DATA_FORMAT_BASE64)) {
-        write_data = g_base64_decode(data, &write_count);
+        write_data = qbase64_decode(data, -1,
+                                    &write_count,
+                                    errp);
+        if (!write_data) {
+            return;
+        }
     } else {
         write_data = (uint8_t *)data;
         write_count = strlen(data);
qmp-commands.hx +0 -2
index 9d8b42f..1b47cdd 100644
@@ -512,8 +512,6 @@ Arguments:
 - "data": data to write (json-string)
 - "format": data format (json-string, optional)
           - Possible values: "utf8" (default), "base64"
-            Bug: invalid base64 is currently not rejected.
-            Whitespace *is* invalid.
 
 Example:
 
Signature · 1 line
2.5.0
Reviewed-by on v2 2/5
#5 lore raw

Re: [Qemu-devel] [PATCH v2 2/5] qemu-char: convert to use error checked base64 decode

Quoted 10 lines · Daniel P. Berrange

Switch from using g_base64_decode over to qbase64_decode in order to get error checking of the base64 input data.

---
 qapi-schema.json | 2 --
 qemu-char.c      | 8 +++++++-
 qmp-commands.hx  | 2 --
 3 files changed, 7 insertions(+), 5 deletions(-)
Signature · 2 lines
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org
#6 lore raw

PATCH v2 3/5 qga: convert to use error checked base64 decode

In reply to Daniel P. Berrange

Switch from using g_base64_decode over to qbase64_decode in order to get error checking of the base64 input data.

---
 qga/commands-posix.c | 11 +++++++++--
 qga/commands-win32.c | 11 +++++++++--
 qga/commands.c       | 13 ++++++++++++-
 3 files changed, 30 insertions(+), 5 deletions(-)
qga/commands-posix.c +9 -2
index 0ebd473..e495cc2 100644
@@ -29,6 +29,7 @@
 #include "qemu/queue.h"
 #include "qemu/host-utils.h"
 #include "qemu/sockets.h"
+#include "qemu/base64.h"
 
 #ifndef CONFIG_HAS_ENVIRON
 #ifdef __APPLE__
@@ -496,7 +497,10 @@ GuestFileWrite *qmp_guest_file_write(int64_t handle, const char *buf_b64,
     }
 
     fh = gfh->fh;
-    buf = g_base64_decode(buf_b64, &buf_len);
+    buf = qbase64_decode(buf_b64, -1, &buf_len, errp);
+    if (!buf) {
+        return NULL;
+    }
 
     if (!has_count) {
         count = buf_len;
@@ -1909,7 +1913,10 @@ void qmp_guest_set_user_password(const char *username,
     char *chpasswddata = NULL;
     size_t chpasswdlen;
 
-    rawpasswddata = (char *)g_base64_decode(password, &rawpasswdlen);
+    rawpasswddata = (char *)qbase64_decode(password, -1, &rawpasswdlen, errp);
+    if (!rawpasswddata) {
+        return;
+    }
     rawpasswddata = g_renew(char, rawpasswddata, rawpasswdlen + 1);
     rawpasswddata[rawpasswdlen] = '\0';
 
qga/commands-win32.c +9 -2
index 41f6dd9..c9e1773 100644
@@ -34,6 +34,7 @@
 #include "qapi/qmp/qerror.h"
 #include "qemu/queue.h"
 #include "qemu/host-utils.h"
+#include "qemu/base64.h"
 
 #ifndef SHTDN_REASON_FLAG_PLANNED
 #define SHTDN_REASON_FLAG_PLANNED 0x80000000
@@ -357,7 +358,10 @@ GuestFileWrite *qmp_guest_file_write(int64_t handle, const char *buf_b64,
         return NULL;
     }
     fh = gfh->fh;
-    buf = g_base64_decode(buf_b64, &buf_len);
+    buf = qbase64_decode(buf_b64, -1, &buf_len, errp);
+    if (!buf) {
+        return NULL;
+    }
 
     if (!has_count) {
         count = buf_len;
@@ -1276,7 +1280,10 @@ void qmp_guest_set_user_password(const char *username,
         return;
     }
 
-    rawpasswddata = (char *)g_base64_decode(password, &rawpasswdlen);
+    rawpasswddata = (char *)qbase64_decode(password, -1, &rawpasswdlen, errp);
+    if (!rawpasswddata) {
+        return;
+    }
     rawpasswddata = g_renew(char, rawpasswddata, rawpasswdlen + 1);
     rawpasswddata[rawpasswdlen] = '\0';
 
qga/commands.c +12 -1
index bb73e7d..58568d8 100644
@@ -14,6 +14,7 @@
 #include "qga/guest-agent-core.h"
 #include "qga-qmp-commands.h"
 #include "qapi/qmp/qerror.h"
+#include "qemu/base64.h"
 
 /* Maximum captured guest-exec out_data/err_data - 16MB */
 #define GUEST_EXEC_MAX_OUTPUT (16*1024*1024)
@@ -393,10 +394,19 @@ GuestExec *qmp_guest_exec(const char *path,
     GIOChannel *in_ch, *out_ch, *err_ch;
     GSpawnFlags flags;
     bool has_output = (has_capture_output && capture_output);
+    uint8_t *input = NULL;
+    size_t ninput = 0;
 
     arglist.value = (char *)path;
     arglist.next = has_arg ? arg : NULL;
 
+    if (has_input_data) {
+        input = qbase64_decode(input_data, -1, &ninput, err);
+        if (!input) {
+            return NULL;
+        }
+    }
+
     argv = guest_exec_get_args(&arglist, true);
     envp = has_env ? guest_exec_get_args(env, false) : NULL;
 
@@ -425,7 +435,8 @@ GuestExec *qmp_guest_exec(const char *path,
     g_child_watch_add(pid, guest_exec_child_watch, gei);
 
     if (has_input_data) {
-        gei->in.data = g_base64_decode(input_data, &gei->in.size);
+        gei->in.data = input;
+        gei->in.size = ninput;
 #ifdef G_OS_WIN32
         in_ch = g_io_channel_win32_new_fd(in_fd);
 #else
Signature · 1 line
2.5.0
Reviewed-by on v2 3/5
#7 lore raw

Re: [Qemu-devel] [PATCH v2 3/5] qga: convert to use error checked base64 decode

Quoted 10 lines · Daniel P. Berrange

Switch from using g_base64_decode over to qbase64_decode in order to get error checking of the base64 input data.

---
 qga/commands-posix.c | 11 +++++++++--
 qga/commands-win32.c | 11 +++++++++--
 qga/commands.c       | 13 ++++++++++++-
 3 files changed, 30 insertions(+), 5 deletions(-)
Signature · 2 lines
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org
#8 lore raw

PATCH v2 4/5 crypto: add QCryptoSecret object class for password/key handling

In reply to Daniel P. Berrange

Introduce a new QCryptoSecret object class which will be used for providing passwords and keys to other objects which need sensitive credentials.

The new object can provide secret values directly as properties, or indirectly via a file. The latter includes support for file descriptor passing syntax on UNIX platforms. Ordinarily passing secret values directly as properties is insecure, since they are visible in process listings, or in log files showing the CLI args / QMP commands. It is possible to use AES-256-CBC to encrypt the secret values though, in which case all that is visible is the ciphertext. For adhoc developer testing though, it is fine to provide the secrets directly without encryption so this is not explicitly forbidden.

The anticipated scenario is that libvirtd will create a random master key per QEMU instance (eg /var/run/libvirt/qemu/$VMNAME.key) and will use that key to encrypt all passwords it provides to QEMU via '-object secret,....'. This avoids the need for libvirt (or other mgmt apps) to worry about file descriptor passing.

It also makes life easier for people who are scripting the management of QEMU, for whom FD passing is significantly more complex.

Providing data inline (insecure, only for adhoc dev testing)

$QEMU -object secret,id=sec0,data=letmein

Providing data indirectly in raw format

printf "letmein" > mypasswd.txt
$QEMU -object secret,id=sec0,file=mypasswd.txt

Providing data indirectly in base64 format

$QEMU -object secret,id=sec0,file=mykey.b64,format=base64

Providing data with encryption

  $QEMU -object secret,id=master0,file=mykey.b64,format=base64 \
        -object secret,id=sec0,data=[base64 ciphertext],\
	           keyid=master0,iv=[base64 IV],format=base64

Note that 'format' here refers to the format of the ciphertext data. The decrypted data must always be in raw byte format.

More examples are shown in the updated docs.

---
 crypto/Makefile.objs       |   1 +
 crypto/secret.c            | 540 +++++++++++++++++++++++++++++++++++++++++++++
 include/crypto/secret.h    | 148 +++++++++++++
 qapi/crypto.json           |  14 ++
 qemu-options.hx            |  77 +++++++
 tests/.gitignore           |   1 +
 tests/Makefile             |   2 +
 tests/test-crypto-secret.c | 446 +++++++++++++++++++++++++++++++++++++
 8 files changed, 1229 insertions(+)
 create mode 100644 crypto/secret.c
 create mode 100644 include/crypto/secret.h
 create mode 100644 tests/test-crypto-secret.c
crypto/Makefile.objs +1 -0
index b2a0e0b..a3135f1 100644
@@ -7,6 +7,7 @@ crypto-obj-y += tlscreds.o
 crypto-obj-y += tlscredsanon.o
 crypto-obj-y += tlscredsx509.o
 crypto-obj-y += tlssession.o
+crypto-obj-y += secret.o
 
 # Let the userspace emulators avoid linking gnutls/etc
 crypto-aes-obj-y = aes.o
crypto/secret.c new file +540 -0

540 lines not on this page · raw message

include/crypto/secret.h new file +148 -0

148 lines not on this page · raw message

qapi/crypto.json +14 -0
index b058b14..4012659 100644
@@ -19,3 +19,17 @@
 { 'enum': 'QCryptoTLSCredsEndpoint',
   'prefix': 'QCRYPTO_TLS_CREDS_ENDPOINT',
   'data': ['client', 'server']}
+
+
+##
+# QCryptoSecretFormat:
+#
+# The data format that the secret is provided in
+#
+# @raw: raw bytes. When encoded in JSON only valid UTF-8 sequences can be used
+# @base64: arbitrary base64 encoded binary data
+# Since: 2.6
+##
+{ 'enum': 'QCryptoSecretFormat',
+  'prefix': 'QCRYPTO_SECRET_FORMAT',
+  'data': ['raw', 'base64']}
qemu-options.hx +77 -0

83 lines not on this page · raw message

tests/.gitignore +1 -0
index 00a7fed..dbe7c06 100644
@@ -14,6 +14,7 @@ test-blockjob-txn
 test-coroutine
 test-crypto-cipher
 test-crypto-hash
+test-crypto-secret
 test-crypto-tlscredsx509
 test-crypto-tlscredsx509-work/
 test-crypto-tlscredsx509-certs/
tests/Makefile +2 -0
index 0e1289f..3f22c68 100644
@@ -80,6 +80,7 @@ check-unit-y += tests/test-write-threshold$(EXESUF)
 gcov-files-test-write-threshold-y = block/write-threshold.c
 check-unit-$(CONFIG_GNUTLS_HASH) += tests/test-crypto-hash$(EXESUF)
 check-unit-y += tests/test-crypto-cipher$(EXESUF)
+check-unit-y += tests/test-crypto-secret$(EXESUF)
 check-unit-$(CONFIG_GNUTLS) += tests/test-crypto-tlscredsx509$(EXESUF)
 check-unit-$(CONFIG_GNUTLS) += tests/test-crypto-tlssession$(EXESUF)
 check-unit-$(CONFIG_LINUX) += tests/test-qga$(EXESUF)
@@ -461,6 +462,7 @@ tests/test-mul64$(EXESUF): tests/test-mul64.o $(test-util-obj-y)
 tests/test-bitops$(EXESUF): tests/test-bitops.o $(test-util-obj-y)
 tests/test-crypto-hash$(EXESUF): tests/test-crypto-hash.o $(test-crypto-obj-y)
 tests/test-crypto-cipher$(EXESUF): tests/test-crypto-cipher.o $(test-crypto-obj-y)
+tests/test-crypto-secret$(EXESUF): tests/test-crypto-secret.o $(test-crypto-obj-y)
 
 tests/crypto-tls-x509-helpers.o-cflags := $(TASN1_CFLAGS)
 tests/crypto-tls-x509-helpers.o-libs := $(TASN1_LIBS)
tests/test-crypto-secret.c new file +446 -0

446 lines not on this page · raw message

Signature · 1 line
2.5.0
#9 lore raw

Re: [Qemu-devel] [PATCH v2 4/5] crypto: add QCryptoSecret object class for password/key handling

Quoted 4 lines · Daniel P. Berrange

Introduce a new QCryptoSecret object class which will be used for providing passwords and keys to other objects which need sensitive credentials.

Quoted 8 lines

The new object can provide secret values directly as properties, or indirectly via a file. The latter includes support for file descriptor passing syntax on UNIX platforms. Ordinarily passing secret values directly as properties is insecure, since they are visible in process listings, or in log files showing the CLI args / QMP commands. It is possible to use AES-256-CBC to encrypt the secret values though, in which case all that is visible is the ciphertext. For adhoc developer testing though,

dictionary.com says it's always 'ad hoc' (since it is literally two Latin words), even though the two are always spoken together in English.
And although 'ad-hoc' is starting to gain some traction in modern
usage, squashing it to 'adhoc' just feels like a typo.

Quoted 9 lines

it is fine to provide the secrets directly without encryption so this is not explicitly forbidden.

The anticipated scenario is that libvirtd will create a random master key per QEMU instance (eg /var/run/libvirt/qemu/$VMNAME.key) and will use that key to encrypt all passwords it provides to QEMU via '-object secret,....'. This avoids the need for libvirt (or other mgmt apps) to worry about file descriptor passing.

Quoted 5 lines

It also makes life easier for people who are scripting the management of QEMU, for whom FD passing is significantly more complex.

Providing data inline (insecure, only for adhoc dev testing)

Of course, if we touch it up in one place, you should use consistent spelling throughout :)

Quoted 34 lines

$QEMU -object secret,id=sec0,data=letmein

Providing data indirectly in raw format

printf "letmein" > mypasswd.txt
$QEMU -object secret,id=sec0,file=mypasswd.txt

Providing data indirectly in base64 format

$QEMU -object secret,id=sec0,file=mykey.b64,format=base64

Providing data with encryption

  $QEMU -object secret,id=master0,file=mykey.b64,format=base64 \
        -object secret,id=sec0,data=[base64 ciphertext],\
	           keyid=master0,iv=[base64 IV],format=base64

Note that 'format' here refers to the format of the ciphertext data. The decrypted data must always be in raw byte format.

More examples are shown in the updated docs.

---

crypto/secret.c new file +8 -0
+static void
+qcrypto_secret_load_data(QCryptoSecret *secret,
+                         uint8_t **output,
+                         size_t *outputlen,
+                         Error **errp)
+{
+    int fd;
+    char *data = NULL;

crypto/secret.c line 45

Quoted 10 lines
crypto/secret.c new file +10 -0
+    size_t offset = 0;
+    size_t length = 0;
+
+    *output = NULL;
+    *outputlen = 0;
+
+    if (secret->file) {
+        if (secret->data) {
+            error_setg(errp,
+                       "'file' and 'data' are mutually exclusive");

Is it worth trying to use a qapi flat union to make the mutual exclusion inherent in the type definition, rather than something we have to enforce manually? (I've got more experience with qapi than with Object, so my question may be nonsensical)

crypto/secret.c line 51

Quoted 6 lines
crypto/secret.c new file +6 -0
+            return;
+        }
+        fd = qemu_open(secret->file, O_RDONLY);
+        if (fd < 0) {
+            error_setg_errno(errp, errno,
+                             "Unable to open %s", secret->file);

Using error_setg_file_open() makes for a consistent message on open() failure.

Quoted 4 lines
crypto/secret.c new file +4 -0
+            return;
+        }
+        while (length < (1024 * 1024)) { /* Limit secrets to 1 MB */
+            if ((length - offset) < 1024) {

crypto/secret.c line 65

Quoted 10 lines
crypto/secret.c new file +10 -0
+                length += 1024;
+                data = g_renew(char, data, length);
+            }
+            ssize_t ret = read(fd, data + offset, length - offset);
+            if (ret == 0) {
+                break;
+            }
+            if (ret < 0) {
+                error_setg_errno(errp, errno,
+                                 "Unable to read from %s", secret->file);

Does glib have a convenience function for reading contents of a file?

Quoted 51 lines
crypto/secret.c new file +51 -0
+                close(fd);
+                g_free(data);
+                return;
+            }
+            offset += ret;
+        }
+        close(fd);
+        if (offset) {
+            /* Even though data is raw 8-bit, so may contain
+             * arbitrary NULs, ensure it is explicitly NUL
+             * terminated */
+            *output = g_renew(uint8_t, data, offset + 1);
+            (*output)[offset] = '\0';
+            *outputlen = offset;
+        } else {
+            error_setg(errp, "Secret file %s is empty",
+                       secret->file);
+            g_free(data);
+        }
+    } else if (secret->data) {
+        *outputlen = strlen(secret->data);
+        *output = g_new(uint8_t, *outputlen + 1);
+        memcpy(*output, secret->data, *outputlen + 1);
+    } else {
+        error_setg(errp, "Either 'file' or 'data' must be provided");
+    }
+}
+
+
+static void qcrypto_secret_decrypt(QCryptoSecret *secret,
+                                   const uint8_t *input,
+                                   size_t inputlen,
+                                   uint8_t **output,
+                                   size_t *outputlen,
+                                   Error **errp)
+{
+    uint8_t *key = NULL, *ciphertext = NULL, *iv = NULL;
+    size_t keylen, ciphertextlen, ivlen;
+    QCryptoCipher *aes = NULL;
+    uint8_t *plaintext = NULL;
+
+    *output = NULL;
+    *outputlen = 0;
+
+    if (qcrypto_secret_lookup(secret->keyid,
+                              &key, &keylen,
+                              errp) < 0) {
+        goto cleanup;
+    }
+
+    if (keylen != 32) {

crypto/secret.c line 126

Quoted 10 lines
crypto/secret.c new file +10 -0
+        error_setg(errp, "Key should be 32 bytes in length");
+        goto cleanup;
+    }
+
+    if (!secret->iv) {
+        error_setg(errp, "IV is required to decrypt secret");
+        goto cleanup;
+    }
+
+    iv = (uint8_t *)g_base64_decode(secret->iv, &ivlen);

Shouldn't this be using qbase64_decode()?

Quoted 11 lines

+++ b/include/crypto/secret.h

+/**
+ * QCryptoSecret:
+ *
+ * The QCryptoSecret object provides storage of secrets,
+ * which may be user passwords, encryption keys or any
+ * other kind of sensitive data that is represented as
+ * a sequence of bytes.
+ *

Quoted 2 lines

+ * Providing data directly, insecurely (suitable for
+ * adhoc developer testing only)

Another instance.

Quoted 9 lines

+++ b/qapi/crypto.json

qapi/crypto.json +4 -0
@@ -19,3 +19,17 @@
 { 'enum': 'QCryptoTLSCredsEndpoint',
   'prefix': 'QCRYPTO_TLS_CREDS_ENDPOINT',
   'data': ['client', 'server']}
+
+
+##
+# QCryptoSecretFormat:

qapi/crypto.json line 35

Quoted 10 lines
qapi/crypto.json +10 -0
+#
+# The data format that the secret is provided in
+#
+# @raw: raw bytes. When encoded in JSON only valid UTF-8 sequences can be used
+# @base64: arbitrary base64 encoded binary data
+# Since: 2.6
+##
+{ 'enum': 'QCryptoSecretFormat',
+  'prefix': 'QCRYPTO_SECRET_FORMAT',
+  'data': ['raw', 'base64']}

This part is fine.

qemu-options.hx Dump the network traffic on netdev @var{dev} to the file specified by · line 3684

Quoted 13 lines
qemu-options.hx +5 -0
index 0eea4ee..41c0e59 100644
@@ -3677,6 +3677,83 @@ Dump the network traffic on netdev @var{dev} to the file specified by
 The file format is libpcap, so it can be analyzed with tools such as tcpdump
 or Wireshark.
 
+@item -object secret,id=@var{id},data=@var{string},format=@var{raw|base64}[,keyid=@var{secretid},iv=@var{string}]
+@item -object secret,id=@var{id},file=@var{filename},format=@var{raw|base64}[,keyid=@var{secretid},iv=@var{string}]
+
+Defines a secret to store a password, encryption key, or some other sensitive
+data. The senstive data can either be passed directly via the @var{data}

s/senstive/sensitive/

qemu-options.hx Dump the network traffic on netdev @var{dev} to the file specified by · line 3687

Quoted 3 lines
qemu-options.hx +3 -0
+parameter, or indirectly via the @var{file} parameter. Using the @var{data}
+parameter is insecure unless the sensitive data is encrypted.
+

Overall looks fairly reasonable.

Signature · 2 lines
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org
#10 lore raw

Re: [Qemu-devel] [PATCH v2 4/5] crypto: add QCryptoSecret object class for password/key handling

Quoted 15 lines · Eric Blake
Quoted 15 lines
crypto/secret.c new file +15 -0
+static void
+qcrypto_secret_load_data(QCryptoSecret *secret,
+                         uint8_t **output,
+                         size_t *outputlen,
+                         Error **errp)
+{
+    int fd;
+    char *data = NULL;
+    size_t offset = 0;
+    size_t length = 0;
+
+    *output = NULL;
+    *outputlen = 0;
+
+    if (secret->file) {
Quoted 8 lines

crypto/secret.c line 45

Quoted 3 lines
crypto/secret.c new file +3 -0
+        if (secret->data) {
+            error_setg(errp,
+                       "'file' and 'data' are mutually exclusive");

Is it worth trying to use a qapi flat union to make the mutual exclusion inherent in the type definition, rather than something we have to enforce manually? (I've got more experience with qapi than with Object, so my question may be nonsensical)

Not ensure sure how you'd wire up a qapi type to a QOM property. It is probably possible in some manner, but I not sure its particularly compelling to do it for this.

Quoted 10 lines

crypto/secret.c line 51

Quoted 7 lines
crypto/secret.c new file +6 -0
+            return;
+        }
+        fd = qemu_open(secret->file, O_RDONLY);
+        if (fd < 0) {
+            error_setg_errno(errp, errno,
+                             "Unable to open %s", secret->file);

Using error_setg_file_open() makes for a consistent message on open() failure.

Yep

Quoted 16 lines
Quoted 5 lines
crypto/secret.c new file +4 -0
+            return;
+        }
+        while (length < (1024 * 1024)) { /* Limit secrets to 1 MB */
+            if ((length - offset) < 1024) {

crypto/secret.c line 65

Quoted 10 lines
crypto/secret.c new file +10 -0
+                length += 1024;
+                data = g_renew(char, data, length);
+            }
+            ssize_t ret = read(fd, data + offset, length - offset);
+            if (ret == 0) {
+                break;
+            }
+            if (ret < 0) {
+                error_setg_errno(errp, errno,
+                                 "Unable to read from %s", secret->file);
Quoted 1 line

Does glib have a convenience function for reading contents of a file?

Of course, I completely forgot about g_file_get_contents() which works just fine.

Quoted 26 lines
Quoted 26 lines
crypto/secret.c new file +26 -0
+static void qcrypto_secret_decrypt(QCryptoSecret *secret,
+                                   const uint8_t *input,
+                                   size_t inputlen,
+                                   uint8_t **output,
+                                   size_t *outputlen,
+                                   Error **errp)
+{
+    uint8_t *key = NULL, *ciphertext = NULL, *iv = NULL;
+    size_t keylen, ciphertextlen, ivlen;
+    QCryptoCipher *aes = NULL;
+    uint8_t *plaintext = NULL;
+
+    *output = NULL;
+    *outputlen = 0;
+
+    if (qcrypto_secret_lookup(secret->keyid,
+                              &key, &keylen,
+                              errp) < 0) {
+        goto cleanup;
+    }
+
+    if (keylen != 32) {
+        error_setg(errp, "Key should be 32 bytes in length");
+        goto cleanup;
+    }
+
Quoted 8 lines

crypto/secret.c line 126

Quoted 6 lines
crypto/secret.c new file +6 -0
+    if (!secret->iv) {
+        error_setg(errp, "IV is required to decrypt secret");
+        goto cleanup;
+    }
+
+    iv = (uint8_t *)g_base64_decode(secret->iv, &ivlen);

Shouldn't this be using qbase64_decode()?

Yeah, it really should

Regards,
Daniel

Signature · 4 lines
|: http://berrange.com      -o-    http://www.flickr.com/photos/dberrange/ :|
|: http://libvirt.org              -o-             http://virt-manager.org :|
|: http://autobuild.org       -o-         http://search.cpan.org/~danberr/ :|
|: http://entangle-photo.org       -o-       http://live.gnome.org/gtk-vnc :|
#11 lore raw

PATCH v2 5/5 crypto: add support for loading encrypted x509 keys

In reply to Daniel P. Berrange

Make use of the QCryptoSecret object to support loading of encrypted x509 keys. The optional 'passwordid' parameter to the tls-creds-x509 object type, provides the ID of a secret object instance that holds the decryption password for the PEM file.

 # printf "123456" > mypasswd.txt
 # $QEMU \
    -object secret,id=sec0,filename=mypasswd.txt \
    -object tls-creds-x509,passwordid=sec0,id=creds0,\
            dir=/home/berrange/.pki/qemu,endpoint=server \
    -vnc :1,tls-creds=creds0

This requires QEMU to be linked to GNUTLS >= 3.1.11. If GNUTLS is too old an error will be reported if an attempt is made to pass a decryption password.

---
 crypto/tlscredsx509.c         | 47 +++++++++++++++++++++++++++++++++++++++++++
 include/crypto/tlscredsx509.h |  1 +
 qemu-options.hx               |  8 +++++++-
 3 files changed, 55 insertions(+), 1 deletion(-)
crypto/tlscredsx509.c +47 -0
index d080deb..00bb299 100644
@@ -20,6 +20,7 @@
 
 #include "crypto/tlscredsx509.h"
 #include "crypto/tlscredspriv.h"
+#include "crypto/secret.h"
 #include "qom/object_interfaces.h"
 #include "trace.h"
 
@@ -606,9 +607,29 @@ qcrypto_tls_creds_x509_load(QCryptoTLSCredsX509 *creds,
     }
 
     if (cert != NULL && key != NULL) {
+#if GNUTLS_VERSION_NUMBER >= 0x030111
+        char *password = NULL;
+        if (creds->passwordid) {
+            password = qcrypto_secret_lookup_as_utf8(creds->passwordid,
+                                                     errp);
+            if (!password) {
+                goto cleanup;
+            }
+        }
+        ret = gnutls_certificate_set_x509_key_file2(creds->data,
+                                                    cert, key,
+                                                    GNUTLS_X509_FMT_PEM,
+                                                    password,
+                                                    0);
+#else /* GNUTLS_VERSION_NUMBER < 0x030111 */
+        if (creds->passwordid) {
+            error_setg(errp, "PKCS8 decryption requires GNUTLS >= 3.1.11");
+            goto cleanup;
+        }
         ret = gnutls_certificate_set_x509_key_file(creds->data,
                                                    cert, key,
                                                    GNUTLS_X509_FMT_PEM);
+#endif /* GNUTLS_VERSION_NUMBER < 0x030111 */
         if (ret < 0) {
             error_setg(errp, "Cannot load certificate '%s' & key '%s': %s",
                        cert, key, gnutls_strerror(ret));
@@ -736,6 +757,27 @@ qcrypto_tls_creds_x509_prop_set_sanity(Object *obj,
 }
 
 
+static void
+qcrypto_tls_creds_x509_prop_set_passwordid(Object *obj,
+                                           const char *value,
+                                           Error **errp G_GNUC_UNUSED)
+{
+    QCryptoTLSCredsX509 *creds = QCRYPTO_TLS_CREDS_X509(obj);
+
+    creds->passwordid = g_strdup(value);
+}
+
+
+static char *
+qcrypto_tls_creds_x509_prop_get_passwordid(Object *obj,
+                                           Error **errp G_GNUC_UNUSED)
+{
+    QCryptoTLSCredsX509 *creds = QCRYPTO_TLS_CREDS_X509(obj);
+
+    return g_strdup(creds->passwordid);
+}
+
+
 static bool
 qcrypto_tls_creds_x509_prop_get_sanity(Object *obj,
                                        Error **errp G_GNUC_UNUSED)
@@ -768,6 +810,10 @@ qcrypto_tls_creds_x509_init(Object *obj)
                              qcrypto_tls_creds_x509_prop_get_sanity,
                              qcrypto_tls_creds_x509_prop_set_sanity,
                              NULL);
+    object_property_add_str(obj, "passwordid",
+                            qcrypto_tls_creds_x509_prop_get_passwordid,
+                            qcrypto_tls_creds_x509_prop_set_passwordid,
+                            NULL);
 }
 
 
@@ -776,6 +822,7 @@ qcrypto_tls_creds_x509_finalize(Object *obj)
 {
     QCryptoTLSCredsX509 *creds = QCRYPTO_TLS_CREDS_X509(obj);
 
+    g_free(creds->passwordid);
     qcrypto_tls_creds_x509_unload(creds);
 }
 
include/crypto/tlscredsx509.h +1 -0
index b9785fd..25796d7 100644
@@ -101,6 +101,7 @@ struct QCryptoTLSCredsX509 {
     gnutls_certificate_credentials_t data;
 #endif
     bool sanityCheck;
+    char *passwordid;
 };
 
 
qemu-options.hx +7 -1
index 41c0e59..38fd8a8 100644
@@ -3626,7 +3626,7 @@ expensive operation that consumes random pool entropy, so it is
 recommended that a persistent set of parameters be generated
 upfront and saved.
 
-@item -object tls-creds-x509,id=@var{id},endpoint=@var{endpoint},dir=@var{/path/to/cred/dir},verify-peer=@var{on|off}
+@item -object tls-creds-x509,id=@var{id},endpoint=@var{endpoint},dir=@var{/path/to/cred/dir},verify-peer=@var{on|off},passwordid=@var{id}
 
 Creates a TLS anonymous credentials object, which can be used to provide
 TLS support on network backends. The @option{id} parameter is a unique
@@ -3653,6 +3653,12 @@ in PEM format, in filenames @var{ca-cert.pem}, @var{ca-crl.pem} (optional),
 @var{server-cert.pem} (only servers), @var{server-key.pem} (only servers),
 @var{client-cert.pem} (only clients), and @var{client-key.pem} (only clients).
 
+For the @var{server-key.pem} and @var{client-key.pem} files which
+contain sensitive private keys, it is possible to use an encrypted
+version by providing the @var{passwordid} parameter. This provides
+the ID of a previously created @code{secret} object containing the
+password for decryption.
+
 @item -object filter-buffer,id=@var{id},netdev=@var{netdevid},interval=@var{t}[,queue=@var{all|rx|tx}]
 
 Interval @var{t} can't be 0, this filter batches the packet delivery: all
Signature · 1 line
2.5.0
Reviewed-by on v2 5/5
#12 lore raw

Re: [Qemu-devel] [PATCH v2 5/5] crypto: add support for loading encrypted x509 keys

Quoted 17 lines · Daniel P. Berrange

Make use of the QCryptoSecret object to support loading of encrypted x509 keys. The optional 'passwordid' parameter to the tls-creds-x509 object type, provides the ID of a secret object instance that holds the decryption password for the PEM file.

 # printf "123456" > mypasswd.txt
 # $QEMU \
    -object secret,id=sec0,filename=mypasswd.txt \
    -object tls-creds-x509,passwordid=sec0,id=creds0,\
            dir=/home/berrange/.pki/qemu,endpoint=server \
    -vnc :1,tls-creds=creds0

This requires QEMU to be linked to GNUTLS >= 3.1.11. If GNUTLS is too old an error will be reported if an attempt is made to pass a decryption password.

Quoted 6 lines
---
 crypto/tlscredsx509.c         | 47 +++++++++++++++++++++++++++++++++++++++++++
 include/crypto/tlscredsx509.h |  1 +
 qemu-options.hx               |  8 +++++++-
 3 files changed, 55 insertions(+), 1 deletion(-)
Signature · 2 lines
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org